Skip to main content

emit_swiglu

logicaffeine_compile::codegen_cuda::fused_row

Function emit_swiglu 

Source 
pub fn emit_swiglu(spec: &SwiGluSpec, knobs: &CudaKnobs) -> String
Expand description

Emit a self-contained CUDA program for a recognized SwiGLU map — y[i] = silu(gate[i]) * up[i], one thread per element, self-contained host (deterministic fills + Σ Y checksum). Chip-stamped like the fused-row.